IN AN EARLIER ARTICLE [5] in this journal, an attempt was made to reformulate Sharpe's single index model as a block recursive system. This block recursive system can incorporate simultaneity in a system with a large number of securities without creating the tremendous estimation difficulties associated with a full simultaneous equation system. In order to use a block recursive system, however, the securities must be grouped. The purpose of this extension is to discuss an alternative grouping technique, i.e., cluster analysis rather than factor analysis.' Cluster analysis has been used sparingly in the finance area. Elton and Gruber [2] have utilized the tool as a method of creating homogeneous groups, while Farrell [3, 4] has used cluster analysis to analyze covariation of returns in order to determine homogeneous stock groupings. In current study, a cluster analysis could have been performed on such variables as size of firm, earnings growth rates, debt-to-equity ratios, etc. Instead, the research uses a cluster analysis of intersecurity correlations in order to be consistent with portfolio considerations. The resultant security clusters have the stock returns within each group highly correlated, while the stock returns between different groups have fairly low correlations. TABLE 1